In [1]:
import prody as pr
import py3Dmol
import tempfile
import pandas as pd
from PlotUtils import *
from scipy.ndimage import binary_dilation
import matplotlib.patches as mpatches
In [2]:
fold_pair = '2lv1A_6lniA'
In [3]:
plot_tool = PlotTool(folder='/Users/steveabecassis/Desktop/Pipeline', fold_pair=fold_pair)

@> 2254 atoms and 20 coordinate set(s) were parsed in 0.36s.

@> 492 atoms and 1 coordinate set(s) were parsed in 0.01s.
In [4]:
from IPython.display import display, HTML
from Bio import PDB
from Bio import pairwise2
from Bio.PDB import Superimposer
import py3Dmol
import ipywidgets as widgets



def get_seq_from_structure(structure):
    return ''.join([PDB.Polypeptide.three_to_one(res.resname) 
                    for res in structure.get_residues() 
                    if PDB.Polypeptide.is_aa(res)])

def get_aligned_atoms(structure1, structure2):
    seq1 = get_seq_from_structure(structure1)
    seq2 = get_seq_from_structure(structure2)
    
    alignment = pairwise2.align.globalxx(seq1, seq2)[0]
    
    aligned_atoms1 = []
    aligned_atoms2 = []
    
    seq1_index = 0
    seq2_index = 0
    
    for a1, a2 in zip(alignment[0], alignment[1]):
        if a1 != '-' and a2 != '-':
            res1 = list(structure1.get_residues())[seq1_index]
            res2 = list(structure2.get_residues())[seq2_index]
            if 'CA' in res1 and 'CA' in res2:
                aligned_atoms1.append(res1['CA'])
                aligned_atoms2.append(res2['CA'])
        
        if a1 != '-':
            seq1_index += 1
        if a2 != '-':
            seq2_index += 1
    
    return aligned_atoms1, aligned_atoms2

def toggle_visibility(change):
    if change['name'] == 'value':
        if change['owner'] == toggle1:
            view.setStyle({'model': 0}, {'cartoon': {'color': 'red'}} if change['new'] else {'cartoon': {'hidden': True}})
        elif change['owner'] == toggle2:
            view.setStyle({'model': 1}, {'cartoon': {'color': 'blue'}} if change['new'] else {'cartoon': {'hidden': True}})
    view.render()

def visualize_alignement_structure(PATH_PDB1,PATH_PDB2):
# Set up PDB parser
    parser = PDB.PDBParser()

    # Load structures
    structure1 = parser.get_structure("protein1", PATH_PDB1)
    structure2 = parser.get_structure("protein2", PATH_PDB2)

    # Get aligned atoms
    aligned_atoms1, aligned_atoms2 = get_aligned_atoms(structure1, structure2)

    # Align structures
    super_imposer = Superimposer()
    super_imposer.set_atoms(aligned_atoms1, aligned_atoms2)
    super_imposer.apply(structure2.get_atoms())

    # Save aligned structure
    io = PDB.PDBIO()
    io.set_structure(structure2)
    io.save("aligned_protein2.pdb")

    # Create py3Dmol view
    view = py3Dmol.view(width=800, height=600)

    # Add structures to the view
    with open(pdb_file2_, 'r') as f:
        view.addModel(f.read(), 'pdb')
    with open("aligned_protein2.pdb", 'r') as f:
        view.addModel(f.read(), 'pdb')

    # Set styles
    view.setStyle({'model': 0}, {'cartoon': {'color': 'red'}})
    view.setStyle({'model': 1}, {'cartoon': {'color': 'blue'}})

    # Zoom to fit
    view.zoomTo()

    # Create toggle buttons
    toggle1 = widgets.ToggleButton(description='Toggle Protein 1', value=True)
    toggle2 = widgets.ToggleButton(description='Toggle Protein 2', value=True)



    toggle1.observe(toggle_visibility)
    toggle2.observe(toggle_visibility)

    # Display the view and toggle buttons
    display(HTML(view._make_html()))
    display(widgets.HBox([toggle1, toggle2]))
from Bio.PDB import PDBParser
from Bio import PDB
from Bio import pairwise2

def extract_protein_sequence(pdb_file):
    parser = PDBParser()
    structure = parser.get_structure("protein_structure", pdb_file)

    residue_sequence = ""
    flag = 0
    # Iterate through the structure and extract the residue sequence
    for model in structure:
        for chain in model:
            if flag!=0:
                break
            flag +=1
            for residue in chain:
                if PDB.is_aa(residue):
                    residue_sequence += PDB.Polypeptide.three_to_one(residue.get_resname())
    return residue_sequence



def get_align_indexes(seqA,seqB):
    alignments = pairwise2.align.globalxx(seqA,seqB,one_alignment_only=True)
    best_align = alignments[0]
    seqA = best_align.seqA
    seqB = best_align.seqB
    cursA = 0
    cursB = 0
    seqA_idxs = []
    seqB_idxs = []
    for aa in range(len(seqA)):
        if (seqA[aa] != '-')&(seqB[aa] != '-'):
            seqA_idxs.append(cursA)
            seqB_idxs.append(cursB)
            cursA +=1
            cursB +=1
        if (seqA[aa] == '-')&(seqB[aa] != '-'):
            cursB +=1
        if (seqA[aa] != '-')&(seqB[aa] == '-'):
            cursA +=1
    return seqA_idxs,seqB_idxs


def dilate_with_tolerance(array, tolerance):
    structure = np.ones((2 * tolerance + 1, 2 * tolerance + 1), dtype=int)
    return binary_dilation(array, structure=structure).astype(int)

def load_pred_cmap(fileName):
    cmap = np.load(f'{plot_tool.folder}/{plot_tool.fold_pair}/output_cmap_esm/{fileName}.npy')
    cmap[cmap > 0.4] = 1
    cmap[cmap <= 0.4] = 0
    return cmap



seq_fold1 = extract_protein_sequence(f'{plot_tool.folder}/{plot_tool.fold_pair}/chain_pdb_files/{plot_tool.fold1}.pdb')
seq_fold2 = extract_protein_sequence(f'{plot_tool.folder}/{plot_tool.fold_pair}/chain_pdb_files/{plot_tool.fold2}.pdb')

fold1_idxs,fold2_idxs = get_align_indexes(seq_fold1, seq_fold2)

Original Structure Visualization¶

In [5]:
pdb_file1 = read_pdb_file(f'{plot_tool.folder}/{plot_tool.fold_pair}/chain_pdb_files/{plot_tool.fold1}.pdb')
plot_tool.plot_single_fold(pdb_file1,label=plot_tool.fold1)

pdb_file2 = read_pdb_file(f'{plot_tool.folder}/{plot_tool.fold_pair}/chain_pdb_files/{plot_tool.fold2}.pdb')
plot_tool.plot_single_fold(pdb_file2,label=plot_tool.fold2)

3Dmol.js failed to load for some reason. Please check your browser console for error messages.

3Dmol.js failed to load for some reason. Please check your browser console for error messages.

In [6]:
# plot_tool.plot_fold_alignement()
plot_tool.plot_fold_alignement_(f'{plot_tool.folder}/{plot_tool.fold_pair}/chain_pdb_files/{plot_tool.fold1}.pdb',f'{plot_tool.folder}/{plot_tool.fold_pair}/chain_pdb_files/{plot_tool.fold2}.pdb')

@> 2254 atoms and 20 coordinate set(s) were parsed in 0.36s.

@> 492 atoms and 1 coordinate set(s) were parsed in 0.01s.

6lniA_TEMP
Blue:2lv1A
Red:6lniA

3Dmol.js failed to load for some reason. Please check your browser console for error messages.

Best AlphaFold Prediction¶

In [7]:
df_af = pd.read_csv(f'{plot_tool.folder}/{fold_pair}/Analysis/df_af.csv')
df_af = df_af[df_af.cluster_num != 'Query'].iloc[:,1:-1]
df_af.head(15)
Out[7]:
pdb_file score_pdb1 score_pdb2 Fold cluster_num
5 ShallowMsa_000_unrelaxed_rank_001_alphafold2_p... 0.50247 0.19266 2lv1 000
6 ShallowMsa_000_unrelaxed_rank_002_alphafold2_p... 0.47466 0.13940 2lv1 000
7 ShallowMsa_000_unrelaxed_rank_003_alphafold2_p... 0.45724 0.17159 2lv1 000
8 ShallowMsa_000_unrelaxed_rank_004_alphafold2_p... 0.35562 0.14956 2lv1 000
9 ShallowMsa_000_unrelaxed_rank_005_alphafold2_p... 0.29656 0.16614 2lv1 000
10 ShallowMsa_001_unrelaxed_rank_001_alphafold2_p... 0.59020 0.13666 2lv1 001
11 ShallowMsa_001_unrelaxed_rank_002_alphafold2_p... 0.59049 0.16394 2lv1 001
12 ShallowMsa_001_unrelaxed_rank_003_alphafold2_p... 0.58522 0.14895 2lv1 001
13 ShallowMsa_001_unrelaxed_rank_004_alphafold2_p... 0.58115 0.16312 2lv1 001
14 ShallowMsa_001_unrelaxed_rank_005_alphafold2_p... 0.57299 0.13881 2lv1 001
15 ShallowMsa_002_unrelaxed_rank_001_alphafold2_p... 0.58439 0.15967 2lv1 002
16 ShallowMsa_002_unrelaxed_rank_002_alphafold2_p... 0.58676 0.17172 2lv1 002
17 ShallowMsa_002_unrelaxed_rank_003_alphafold2_p... 0.58428 0.14800 2lv1 002
18 ShallowMsa_002_unrelaxed_rank_004_alphafold2_p... 0.58948 0.17260 2lv1 002
19 ShallowMsa_002_unrelaxed_rank_005_alphafold2_p... 0.58389 0.13942 2lv1 002
In [8]:
df_af[df_af.score_pdb1 > df_af.score_pdb2].head()
Out[8]:
pdb_file score_pdb1 score_pdb2 Fold cluster_num
5 ShallowMsa_000_unrelaxed_rank_001_alphafold2_p... 0.50247 0.19266 2lv1 000
6 ShallowMsa_000_unrelaxed_rank_002_alphafold2_p... 0.47466 0.13940 2lv1 000
7 ShallowMsa_000_unrelaxed_rank_003_alphafold2_p... 0.45724 0.17159 2lv1 000
8 ShallowMsa_000_unrelaxed_rank_004_alphafold2_p... 0.35562 0.14956 2lv1 000
9 ShallowMsa_000_unrelaxed_rank_005_alphafold2_p... 0.29656 0.16614 2lv1 000
In [9]:
try:
    max_af_pdb1 = df_af[df_af.score_pdb1 > df_af.score_pdb2].sort_values(by='score_pdb1',ascending=False).pdb_file.iloc[0]
    max_af_pdb1 = f'{plot_tool.folder}/{fold_pair}/AF_preds/{max_af_pdb1}'
    pdb_file1_ = f'{plot_tool.folder}/{plot_tool.fold_pair}/chain_pdb_files/{plot_tool.fold1}.pdb'
    score = df_af[df_af.score_pdb1 > df_af.score_pdb2].sort_values(by='score_pdb1',ascending=False).score_pdb1.iloc[0]
    plot_tool.align_and_visualize_pdb(max_af_pdb1,pdb_file1_,score)
except:
    print('No max_af_pdb1')

@> 1162 atoms and 1 coordinate set(s) were parsed in 0.03s.

@> 2254 atoms and 20 coordinate set(s) were parsed in 0.35s.

Blue:ShallowMsa_004_unrelaxed_rank_004_alphafold2_ptm_model_3_seed_000
Red:2lv1A

3Dmol.js failed to load for some reason. Please check your browser console for error messages.

In [10]:
# try:
#     plot_tool.plot_single_fold(pdb_file1,label=plot_tool.fold1)
#     max_af_pdb1 = df_af[df_af.score_pdb1 > df_af.score_pdb2].sort_values(by='score_pdb1',ascending=False).pdb_file.iloc[0]
#     max_af_pdb1 = read_pdb_file(f'{plot_tool.folder}/{fold_pair}/AF_preds/{max_af_pdb1}')
#     plot_tool.plot_single_fold(max_af_pdb1,label='MAX AF TM PDB1')
# except Exception as e:
#     print(e)
In [11]:
df_af[df_af.score_pdb1 < df_af.score_pdb2].head()
Out[11]:
pdb_file score_pdb1 score_pdb2 Fold cluster_num
In [12]:
# df_af[df_af.score_pdb2 > df_af.score_pdb1].sort_values(by='score_pdb2',ascending=False)
In [13]:
import py3Dmol
from Bio import PDB, Align
from Bio.PDB import PDBParser, PDBIO
from Bio.PDB.Polypeptide import three_to_one
from io import StringIO
import numpy as np

def visualize_structure_alignment(pdb_file1, pdb_file2, chain1='A', chain2='A'):
    # Parse PDB files
    parser = PDBParser(QUIET=True)
    structure1 = parser.get_structure("protein1", pdb_file1)
    structure2 = parser.get_structure("protein2", pdb_file2)

    # Extract sequences
    seq1 = "".join([three_to_one(res.resname) for res in structure1[0][chain1] if res.id[0] == ' '])
    seq2 = "".join([three_to_one(res.resname) for res in structure2[0][chain2] if res.id[0] == ' '])

    # Perform sequence alignment
    aligner = Align.PairwiseAligner()
    alignments = aligner.align(seq1, seq2)
    best_alignment = next(alignments)

    # Get aligned positions
    aligned_positions = [(i, j) for i, j in zip(best_alignment.path[0], best_alignment.path[1]) 
                         if i != -1 and j != -1]

    # Extract aligned CA atoms
    ca_atoms1 = [atom for atom in structure1[0][chain1].get_atoms() if atom.name == 'CA']
    ca_atoms2 = [atom for atom in structure2[0][chain2].get_atoms() if atom.name == 'CA']
    
    aligned_atoms1 = [ca_atoms1[i] for i, _ in aligned_positions]
    aligned_atoms2 = [ca_atoms2[j] for _, j in aligned_positions]

    # Superimpose structures
    super_imposer = PDB.Superimposer()
    super_imposer.set_atoms(aligned_atoms1, aligned_atoms2)
    super_imposer.apply(structure2.get_atoms())

    # Function to get PDB string
    def get_pdb_string(structure):
        io = StringIO()
        pdb_io = PDBIO()
        pdb_io.set_structure(structure)
        pdb_io.save(io)
        return io.getvalue()

    # Create a py3Dmol view
    view = py3Dmol.view(width=800, height=600)

    # Add the first structure
    view.addModel(get_pdb_string(structure1), 'pdb')
    view.setStyle({'model': -1}, {'cartoon': {'color': 'blue'}})

    # Add the second (aligned) structure
    view.addModel(get_pdb_string(structure2), 'pdb')
    view.setStyle({'model': -1}, {'cartoon': {'color': 'red'}})

    # Zoom to fit the structures
    view.zoomTo()

    # Calculate RMSD
    rmsd = super_imposer.rms

    # Add RMSD and alignment info as labels
    view.addLabel(f"RMSD: {rmsd:.2f} Å", {'position': {'x': -20, 'y': -20}, 'backgroundColor': 'white', 'fontColor': 'black'})
    view.addLabel(f"Aligned: {len(aligned_positions)} of {len(seq1)} (blue) and {len(seq2)} (red) residues", 
                  {'position': {'x': -20, 'y': -40}, 'backgroundColor': 'white', 'fontColor': 'black'})

    return view

# Example usage:
# view = visualize_structure_alignment('protein1.pdb', 'protein2.pdb')
# view.show()
In [14]:
try:
    max_af_pdb2 = df_af[df_af.score_pdb2 > df_af.score_pdb1].sort_values(by='score_pdb2',ascending=False).pdb_file.iloc[0]
    score = df_af[df_af.score_pdb2 > df_af.score_pdb1].sort_values(by='score_pdb2',ascending=False).score_pdb2.iloc[0]
    max_af_pdb2_ = f'{plot_tool.folder}/{fold_pair}/AF_preds/{max_af_pdb2}'
    pdb_file2_ = f'{plot_tool.folder}/{plot_tool.fold_pair}/chain_pdb_files/{plot_tool.fold2}.pdb'
    print(f'red : {plot_tool.fold2}')
    print(f'blue : {max_af_pdb2}')
    #     plot_tool.plot_fold_alignement_(pdb_file2_,max_af_pdb2_)
    visualize_alignement_structure(pdb_file2_,max_af_pdb2_)
#     plot_tool.align_and_visualize_pdb(max_af_pdb2_,pdb_file2_)

except Exception as e:
    print(e)

    

single positional indexer is out-of-bounds

Best EsmFold Prediction¶

In [15]:
df_esmfold_analysis = pd.read_csv(f'{plot_tool.folder}_res/df_esmfold_analysis.csv')
df_esmfold = df_esmfold_analysis[df_esmfold_analysis.fold_pair == fold_pair]
df_esmfold_analysis.head()
Out[15]:
fold_pair fold TMscore_fold1 TMscore_fold2 cluster_num TM_mean_cluster_pdb1 TM_mean_cluster_pdb2 sample_count is_fold_1 is_fold_2 cluster_fold_1 cluster_fold_2
0 1dzlA_5keqF ShallowMsa_000_1.pdb 0.21478 0.22071 000 0.281451 0.2755 10 1 0 1 0
1 1dzlA_5keqF ShallowMsa_000_0.pdb 0.23730 0.24378 000 0.281451 0.2755 10 1 0 1 0
2 1dzlA_5keqF ShallowMsa_000_2.pdb 0.38172 0.36978 000 0.281451 0.2755 10 1 0 1 0
3 1dzlA_5keqF ShallowMsa_000_3.pdb 0.38407 0.38314 000 0.281451 0.2755 10 1 0 1 0
4 1dzlA_5keqF ShallowMsa_000_7.pdb 0.18823 0.17477 000 0.281451 0.2755 10 1 0 1 0
In [16]:
try:
    max_esm_pdb1 = df_esmfold[df_esmfold.TM_mean_cluster_pdb1 > df_esmfold.TM_mean_cluster_pdb2].sort_values(by='TMscore_fold1',ascending=False).fold.iloc[0]
    score = df_esmfold[df_esmfold.TM_mean_cluster_pdb1 > df_esmfold.TM_mean_cluster_pdb2].sort_values(by='TMscore_fold1',ascending=False).TMscore_fold1.iloc[0]

#     max_esm_pdb1_ = read_pdb_file(f'{plot_tool.folder}/{fold_pair}/output_esm_fold/{max_esm_pdb1}')
#     plot_tool.plot_single_fold(max_esm_pdb1_,label='MAX ESM TM PDB1')

    plot_tool.align_and_visualize_pdb(pdb_file1_,f'{plot_tool.folder}/{fold_pair}/output_esm_fold/{max_esm_pdb1}',score)
except Exception as e:
    print(e)
    

@> 2254 atoms and 20 coordinate set(s) were parsed in 0.36s.

@> 1162 atoms and 1 coordinate set(s) were parsed in 0.03s.

Blue:2lv1A
Red:ShallowMsa_008_4

3Dmol.js failed to load for some reason. Please check your browser console for error messages.

In [17]:
try:
    df_esmfold = df_esmfold[df_esmfold.fold.str.contains('ShallowMsa')]
    if len(df_esmfold[df_esmfold.TM_mean_cluster_pdb1 < df_esmfold.TM_mean_cluster_pdb2])>0:
        max_esm_pdb2 = df_esmfold[df_esmfold.TM_mean_cluster_pdb1 < df_esmfold.TM_mean_cluster_pdb2].sort_values(by='TMscore_fold2',ascending=False).fold.iloc[0]
        score = df_esmfold[df_esmfold.TM_mean_cluster_pdb1 < df_esmfold.TM_mean_cluster_pdb2].sort_values(by='TMscore_fold2',ascending=False).TMscore_fold2.iloc[0]
        print(f'red:{plot_tool.fold2}')
        print(f'blue:{max_esm_pdb2}')
        visualize_alignement_structure(f'{plot_tool.folder}/{plot_tool.fold_pair}/chain_pdb_files/{plot_tool.fold2}.pdb',f'{plot_tool.folder}/{fold_pair}/output_esm_fold/{max_esm_pdb2}')

        
except Exception as e:
    print(e)

Contact maps¶

In [18]:
try:
    df_cmap = pd.read_csv(f'{plot_tool.folder}/{fold_pair}/Analysis/new_cmap_res.csv')
    df_cmap['f1Score_1'] = (df_cmap['precision_1']*df_cmap['recall_1'])/(df_cmap['precision_1']+df_cmap['recall_1'])
    df_cmap['f1Score_2'] = (df_cmap['precision_2']*df_cmap['recall_2'])/(df_cmap['precision_2']+df_cmap['recall_2'])

    df_cmap.head()
except Exception as e:
    print(e)
In [19]:
try:
    max_f1Score_pdb1 = df_cmap[df_cmap.f1Score_1 > df_cmap.f1Score_2].sort_values(by='f1Score_1',ascending=False).iloc[0]
except:
    print('No max_f1Score_pdb1')
try:
    max_f1Score_pdb2 = df_cmap[df_cmap.f1Score_1 < df_cmap.f1Score_2].sort_values(by='f1Score_2',ascending=False).iloc[0]
except:
    print('No max_f1Score_pdb2')
try:
    max_precision_pdb1 = df_cmap[df_cmap.precision_1 > df_cmap.precision_2].sort_values(by='precision_1',ascending=False).iloc[0]
except:
    print('No max_precision_pdb1')
try:
    max_precision_pdb2 = df_cmap[df_cmap.precision_1 < df_cmap.precision_2].sort_values(by='precision_2',ascending=False).iloc[0]
except:
    print('No max_precision_pdb2')
try:
    max_recall_pdb1 = df_cmap[df_cmap.recall_1 > df_cmap.recall_2].sort_values(by='recall_1',ascending=False).iloc[0]
except:
    print('No max_recall_pdb1')
try:
    max_recall_pdb2 = df_cmap[df_cmap.recall_1 < df_cmap.recall_2].sort_values(by='recall_2',ascending=False).iloc[0]
except:
    print('No max_recall_pdb2')

No max_precision_pdb2




















In [20]:






tolerance = 2
cmap_pdb1 = plot_tool.get_contact_map_from_pdb(f'{plot_tool.folder}/{plot_tool.fold_pair}/chain_pdb_files/{plot_tool.fold1}.pdb')[fold1_idxs][:, fold1_idxs]
cmap_pdb2 = plot_tool.get_contact_map_from_pdb(f'{plot_tool.folder}/{plot_tool.fold_pair}/chain_pdb_files/{plot_tool.fold2}.pdb')[fold2_idxs][:, fold2_idxs]

dilated_cmap1 = dilate_with_tolerance(cmap_pdb1, tolerance)
dilated_cmap2 = dilate_with_tolerance(cmap_pdb2, tolerance)
overlap_dilated = np.logical_and(dilated_cmap1, dilated_cmap2).astype(int)

try:
    cmap_pred_f1_1 = load_pred_cmap(max_f1Score_pdb1.fileName)[fold1_idxs][:, fold1_idxs]
except:
    print('No max_f1Score_pdb1')
try:
    cmap_pred_f1_2 = load_pred_cmap(max_f1Score_pdb2.fileName)[fold1_idxs][:, fold1_idxs]
except:
    print('No max_f1Score_pdb2')
try:
    cmap_pred_precision_1 = load_pred_cmap(max_precision_pdb1.fileName)[fold1_idxs][:, fold1_idxs]
except:
    print('No cmap_pred_precision_1')
try:
    cmap_pred_precision_2 = load_pred_cmap(max_precision_pdb2.fileName)[fold1_idxs][:, fold1_idxs]
except:
    print('No cmap_pred_precision_2')
    
try:    
    cmap_pred_recall_1 = load_pred_cmap(max_recall_pdb1.fileName)[fold1_idxs][:, fold1_idxs]
except:
    print('No cmap_pred_recall_1')
    
try:
    cmap_pred_recall_2 = load_pred_cmap(max_recall_pdb2.fileName)[fold1_idxs][:, fold1_idxs]
except:
    print('No cmap_pred_recall_2')

only_fold1 = np.logical_and(cmap_pdb1, np.logical_not(overlap_dilated))
only_fold2 = np.logical_and(cmap_pdb2, np.logical_not(overlap_dilated))

























No cmap_pred_precision_2





















In [21]:






try:        
    plt.figure(figsize=(15, 8))# You can adjust the size as needed
    try:
        subplot(1,2,1)
        upper_array = np.clip(np.multiply(dilate_with_tolerance(cmap_pred_f1_1, tolerance), only_fold1),0,1)
        lower_array = np.clip(cmap_pdb1,0,1)
        n = lower_array.shape[0]
        upper_mask     = np.tril(np.ones(n),k=0)[::-1]
        lower_mask     = np.logical_not(upper_mask)[::-1].astype(int)
        overlap_lower  = 0.75*np.multiply(dilated_cmap2,lower_mask).astype(int)
        predicted_map  = 0.75*np.multiply(cmap_pred_f1_1,upper_mask[::-1])
        combined_array = np.clip(np.multiply(lower_array,lower_mask)+ 0.25*predicted_map - overlap_lower,0,1) + np.clip(np.multiply(upper_array,upper_mask[::-1]),0,1)
        mappable1      = imshow(np.clip(combined_array,0,1),origin='lower', cmap='Blues')
        plt.title(f'Orignal {plot_tool.fold2} Vs Best F1 score {round(max_f1Score_pdb2.f1Score_1,2)}')
    except:
        print(f'No plot for f1Score_1')
    try:
        subplot(1,2,2)
        upper_array = np.clip(np.multiply(dilate_with_tolerance(cmap_pred_f1_2, tolerance), only_fold2),0,1)
        lower_array = np.clip(cmap_pdb2,0,1)
        n = lower_array.shape[0]
        upper_mask     = np.tril(np.ones(n),k=0)[::-1]
        lower_mask     = np.logical_not(upper_mask)[::-1].astype(int)
        overlap_lower  = 0.75*np.multiply(dilated_cmap1,lower_mask).astype(int)
        predicted_map  = 0.75*np.multiply(cmap_pred_f1_2,upper_mask[::-1])
        combined_array = np.clip(np.multiply(lower_array,lower_mask)+ 0.25*predicted_map - overlap_lower,0,1) + np.clip(np.multiply(upper_array,upper_mask[::-1]),0,1)
        mappable1      = imshow(np.clip(combined_array,0,1),origin='lower', cmap='Blues')
        plt.title(f'Orignal {plot_tool.fold2} Vs Best F1 score {round(max_f1Score_pdb2.f1Score_2,2)}')
    except:
        print(f'No plot for f1Score_2')
    white_patch = mpatches.Patch(color='lightsteelblue', label='Overlap')
    plt.figlegend(handles=[white_patch], loc='lower center', bbox_to_anchor=(0.5, -0.05), ncol=2, title='Contact map legend')

except Exception as e:
    print(e)
# print('max_f1Score_pdb1:')
# print(max_f1Score_pdb1)
# print('max_f1Score_pdb2:')
# print(max_f1Score_pdb2)
























No description has been provided for this image



















In [22]:






try:        
    plt.figure(figsize=(15, 8))  # You can adjust the size as needed
    subplot(1,2,1)
    try:
        upper_array = np.clip(np.multiply(dilate_with_tolerance(cmap_pred_precision_1, tolerance), only_fold1),0,1)
        lower_array = np.clip(cmap_pdb1,0,1)
        n = lower_array.shape[0]
        upper_mask     = np.tril(np.ones(n),k=0)[::-1]
        lower_mask     = np.logical_not(upper_mask)[::-1].astype(int)
        overlap_lower  = 0.75*np.multiply(dilated_cmap2,lower_mask).astype(int)
        predicted_map  = 0.75*np.multiply(cmap_pred_precision_1,upper_mask[::-1])
        combined_array = np.clip(np.multiply(lower_array,lower_mask)+ 0.25*predicted_map - overlap_lower,0,1) + np.clip(np.multiply(upper_array,upper_mask[::-1]),0,1)
        mappable1      = imshow(np.clip(combined_array,0,1),origin='lower', cmap='Blues')
        plt.title(f'Orignal {plot_tool.fold2} Vs Best Precision score {round(max_precision_pdb1.precision_1,2)}')
    except:
        print(f'No precision_1')
        
    try:
        subplot(1,2,2)
        upper_array = np.clip(np.multiply(dilate_with_tolerance(cmap_pred_precision_2, tolerance), only_fold2),0,1)
        lower_array = np.clip(cmap_pdb2,0,1)
        n = lower_array.shape[0]
        upper_mask     = np.tril(np.ones(n),k=0)[::-1]
        lower_mask     = np.logical_not(upper_mask)[::-1].astype(int)
        overlap_lower  = 0.75*np.multiply(dilated_cmap1,lower_mask).astype(int)
        predicted_map  = 0.75*np.multiply(cmap_pred_precision_2,upper_mask[::-1])
        combined_array = np.clip(np.multiply(lower_array,lower_mask)+ 0.25*predicted_map - overlap_lower,0,1) + np.clip(np.multiply(upper_array,upper_mask[::-1]),0,1)
        mappable1      = imshow(np.clip(combined_array,0,1),origin='lower', cmap='Blues')
        plt.title(f'Orignal {plot_tool.fold2} Vs Best Precision score {round(max_precision_pdb2.precision_2,2)}')
    except:
        print(f'No precision_2')
    white_patch = mpatches.Patch(color='lightsteelblue', label='Overlap')
    plt.figlegend(handles=[white_patch], loc='lower center', bbox_to_anchor=(0.5, -0.05), ncol=2, title='Contact map legend')
except Exception as e:
    print(e)

























No precision_2












No description has been provided for this image



















In [23]:






import matplotlib.pyplot as plt
import numpy as np
import matplotlib.patches as mpatches
from IPython.display import display

try:        
    fig, axs = plt.subplots(1, 2, figsize=(15, 8))  # Create a figure and two subplots

    # Plot for the first subplot
    try:
        upper_array = np.clip(np.multiply(dilate_with_tolerance(cmap_pred_recall_1, tolerance), only_fold1), 0, 1)
        lower_array = np.clip(cmap_pdb1, 0, 1)
        n = lower_array.shape[0]
        upper_mask = np.tril(np.ones(n), k=0)[::-1]
        lower_mask = np.logical_not(upper_mask)[::-1].astype(int)
        overlap_lower = 0.75 * np.multiply(dilated_cmap2, lower_mask).astype(int)
        predicted_map = 0.75 * np.multiply(cmap_pred_recall_1, upper_mask[::-1])
        combined_array = np.clip(np.multiply(lower_array, lower_mask) + 0.25 * predicted_map - overlap_lower, 0, 1) + np.clip(np.multiply(upper_array, upper_mask[::-1]), 0, 1)
        axs[0].imshow(np.clip(combined_array, 0, 1), origin='lower', cmap='Blues')
        axs[0].set_title(f'Orignal {plot_tool.fold2} Vs Best Recall score {round(max_recall_pdb1.recall_1, 2)}')
    except:
        print('No recall_1')
    
    try:
        # Plot for the second subplot
        upper_array = np.clip(np.multiply(dilate_with_tolerance(cmap_pred_recall_2, tolerance), only_fold2), 0, 1)
        lower_array = np.clip(cmap_pdb2, 0, 1)
        n = lower_array.shape[0]
        upper_mask = np.tril(np.ones(n), k=0)[::-1]
        lower_mask = np.logical_not(upper_mask)[::-1].astype(int)
        overlap_lower = 0.75 * np.multiply(dilated_cmap1, lower_mask).astype(int)
        predicted_map = 0.75 * np.multiply(cmap_pred_recall_2, upper_mask[::-1])
        combined_array = np.clip(np.multiply(lower_array, lower_mask) + 0.25 * predicted_map - overlap_lower, 0, 1) + np.clip(np.multiply(upper_array, upper_mask[::-1]), 0, 1)
        axs[1].imshow(np.clip(combined_array, 0, 1), origin='lower', cmap='Blues')
        axs[1].set_title(f'Orignal {plot_tool.fold2} Vs Best Recall score {round(max_recall_pdb2.recall_2, 2)}')
    except:
         print('No recall_2')
    white_patch = mpatches.Patch(color='lightsteelblue', label='Overlap')
    # plt.figlegend(handles=[white_patch], loc='lower center', bbox_to_anchor=(0.5, -0.05), ncol=2, title='Contact map legend')
    plt.show()  # Ensure the plot is displayed

except Exception as e:
    print(f"An error occurred: {e}")
























No description has been provided for this image